Selected Issues in HIV Clinical Trials Judith S. Currier, M.D., MSc Professor of Medicine Division of Infectious Diseases University of California, Los Angeles
Issues Evolving Global and Domestic Epidemic Evaluating new therapies Non-inferiority trials Composite Endpoints Sex differences and clinical trials
Global AIDS Pandemic: UNAIDS 2013
Targets for HIV Therapy Budding Pommier Y, et al. Nat Rev Drug Discov. 2005;4:236-248.
Antiretroviral Drugs 2014 reverse transcriptase inhibitors(12) etravirine (ETV) rilpivirine (RPV protease inhibitors (9) saquinavir (SQV) nucleoside analogues ritonavir (RTV) zidovudine (AZT, ZDV) indinavir (IDV) didanosine (ddi) nelfinavir (NFV) zalcitabine (ddc) amprenavir (APV) stavudine (d4t) lopinavir/r (LPV/r) lamivudine (3TC) fosamprenavir (FPV) abacavir (ABC) integrase inhibitor atazanavir (ATV) emtricitabine (FTC) raltegravir (RAL) tipranavir (TPV) nucleotide analogue Elvitegravir tenofovir (TFV) dolutegravir darunavir (DRV) non-nucleoside analogues fusion inhibitor fuzeon (T20) nevirapine (NVP) entry inhibitor (CCR5) delavirdine (DLV) maraviroc (MV) efavirenz (EFV)
New Formulations/Doses Simplified Regimens Circa 1996 One Pill/Day 2006 efavirenz emtricitabine tenofovir 2014 single tablet regimens Elvitegravir/COBI/FTC/TDF Also RPV/TDF/FTC and DTG/ABC/3TC coming soon
What is the role of newer agents in initial therapy? CASTLE Atazanavir/r vs Lopinavir /r Lancet. 2008;23;372:646-55. ARTEMIS Darunavir/r vs LPV/r Mills A, et al. ICAAC/IDSA 2008. Abstract 1250c. AIDS. 2008 Jul 31;22(12):1389-97. Raltegravir vs Efavirenz STARTMRK Lennox J, et al. ICAAC/IDSA 2008. Abstract 896a
Non-inferiority Design
Objective To show experimental therapy is no worse (in a practical sense) than active control Rule out important or relevant (clinical and statistical) differences with reasonable confidence Underlying implication is that the experimental therapy is thus superior to placebo Cannot be demonstrated with non-significant p-values High p-value similarity Scientific method Absence of evidence is not evidence of absence
Rationale for Non-Inferiority in HIV Trials Experimental therapy often better in other ways Better toxicity profile Less expensive - Favorable adherence profile Important for resource limited settings Less invasive or complicated (better adherence) Shorter treatment duration
Design Assumptions Constancy Efficacy of active control has not changed since showing superiority to placebo Changes could occur with the development of resistance or adjuvant therapies Assay sensitivity Able to detect differences if they exist E.g., poor adherence in both arms results in similarity Composite endpoint (More on this later) Concern for dilution of effect
Selection of the NI Interval Combination of statistical reasoning and clinical judgment Must be smaller than the effect size of active control over placebo Conceptually Maximum treatment difference that is clinically irrelevant Largest treatment difference that is acceptable in order to gain other advantages of the experimental intervention Context dependent Pre-specification important Ideally chosen independent of considerations of study power But sample size is very sensitive to its selection, affecting feasibility Directly impacts study conclusions
Alternative: Estimation-based Design Particularly useful when the acceptable NI margin is not universal Estimate difference between arms with acceptable precision (as measured by the width of a CI)
Non-inferiority Trials and HIV Reviewed 18 recent clinical trials of HIV therapy and evaluated the adherence to standards for reporting Pre-specified NI margins ranged from 10-15% 38% reported ITT results only Half of the studies had non-conclusive results for non-inferiority
Conclusions about non-inferiority should be drawn on the basis of the confidence interval analysis of an appropriate primary endpoint, using the predefined criteria for noninferiority, in both on treatment (OT) and intent to treat (ITT) analyses.
1 Report/justify NI margin 2 Sample size justified 3 Both ITT and AT 4 Report CI of difference 5.1 base conclusions of both ITT or AT 5.2 Restate NI margin in abstract 5.3 Interpret based on primary endpoint 5.4 conclusion matches aim of study- non-inferior or equivalent Parienti BMC Medical Research Methodology 2006, 6:46
Piaggio G et al JAMA, March 8, 2006:1152-1160
New drug MVC compared to standard Noninferiority was defined as the lower bound of this 97.5% CI being above 10%.
Important Lessons: MERIT Lack of precision for entry criteria (sensitivity of CCR5 test) led to wider confidence interval around the difference in response Smaller pre-specified CI for NI (10%) led to rejection of NI MVC was not selected as a preferred initial regimen for treatment of HIV by guidelines
Composite Endpoints Combination of multiple endpoints Used to reduce sample size Usually includes a disease specific endpoint and all cause mortality Interpretation is problematic when outcome is driven by one component Often not reported accurately Overstatement of benefit of treatment
Freemantle N et al. JAMA 2003;289:2554-59
Composite Endpoints in HIV Trials Virologic failure and toxicity A more comprehensive picture of the outcome May be more efficient due to low rates of virologic failure endpoints But does it really capture what we want to know? Important that the severity of toxicity be captured grade 3 and above? One component may precede the other in time
Freemantle JAMA 2003;289:2554-59
Are There Sex Differences in Response to ARV Therapy?
Percent of adults (15+) living with HIV who are female, 1990 2007 70 60 50 Percent 40 female (%) 30 20 10 0 1990 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 062007 Year Sub-Saharan Africa GLOBAL Caribbean Asia E Europe & C Asia Latin America 2007 AIDS epidemic update 5
2007 AIDS epidemic update
Treatment discontinuation or modification greater in women Percentage discontinuing EFV 56.3% Women 38.8% 41.8% 28.3% Men Royal Free Clinic: Women 1.5x more likely to d/c EFV than men (naives) 1 CASTLE (ATV/r vs LPV/r): Women more likely to d/c (21% vs 14%; 29% vs 18%); sex differences in efficacy in ITT 2 Meta-analysis- earlier and greater d/c or change of HCV tx in HIV+ women 5 1Smith. JAIDS 2007; 2Johnson. Gender-based differences in antiretroviral-naïve patients treated with ritonavir-boosted protease inhibitors: results from the CASTLE study through 96 weeks. 12th European AIDS Conference, 2009; 3Bhattacharya. JAIDS Oct 2010
Nicastri. Sex issues in HIV-1 infected persons during HAART: a systematic review. JACC 2007 Higher rates of side effects on HIV therapy in women Adverse event higher in men Adverse event higher in women Ratios (+/- 95% CI) of different adverse events by sex with HAART (2002-07) 1
Limitations of Current Data Most randomized trials of ARV are underpowered to detect sex differences in primary endpoint Would need studies with large sample size over 20% women Need to adjust for age, race/ethnicity As treatments improve, overall response rates are high and differences more difficult to detect Large Cohorts provide the best data currently
J Currier et al., VIII Int Conference on AIDS, Amsterdam 1992, Abstract 4705 Sample Size Calculations to Detect a 50% Reduction in 2-Year Failure Rates Between Women and Men (1-beta=.8, alpha=.025) % accrual of women Total Sample Size 3500 3000 2500 2000 1500 1000 500 10% accrual 15% accrual 20% accrual 0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 Failure Rate in Women
Sample Sizes Required to Detect 2-Fold Differences in Toxicity Rates Between Men and Women Rate in Men Rate in Women Accural of Women Total Sample Size 5% 10% 10% 2353 5% 10% 15% 1689 5% 10% 20% 1367 10% 20% 10% 1095 10% 20% 15% 785 10% 20% 20% 879 20% 40% 10% 466 20% 40% 15% 333 J Currier et al., VIII Int Conference on AIDS, Amsterdam 1992, Abstract 4705
1Sohler et al. Gender Disparities in HIV Health Care Utilization among the Severely Disadvantaged: Can We Determine the Reasons? AIDS Patient Care and STDs, 2009; 2Braitstein. J Women s Health 2008; 3Weiser. JAIDS 2009; 4Ickovics. JAMA 2001; 5Ghebremichael M AIDS Care 2009; 6Cook et al. AIDS 2008 Further barriers to HIV treatment in women- LIFE Sex differences in HIV treatment utilization still exist (in US, not global) 1,2 Lower education levels Lower rates of having health insurance Competing priorities (food 3, childcare, etc.) Depression 4 Substance use 5,6 Domestic violence
Study flow diagram.* Other classification was selected by the investigator as the reason for discontinuation. Older patient taking too many concomitant medications. Currier J et al. Ann Intern Med 2010;153:349-357 2010 by American College of Physicians
Confirmed virologic response in ITT population (top) and population that censored patients who withdrew for reasons other than virologic failure (bottom).virologic response was defined as viral load less than 50 copies/ml, confirmed by 2 consecutive assessments at least 14 days apart. Currier J et al. Ann Intern Med 2010;153:349-357 2010 by American College of Physicians
Confirmed virologic response in ITT population (top) and population that censored patients who withdrew for reasons other than virologic failure (bottom).virologic response was defined as viral load less than 50 copies/ml, confirmed by 2 consecutive assessments at least 14 days apart. The absolute difference in response, based on logistic regression and adjusted for baseline log10 viral load and CD4 cell count, was 9.6 percentage points (95% CI, 19.9 to 0.7%; P 0.067) for intention-to-treat TLOVR and 3.9 percentage points (CI, 13.9 %to 6.0 %; P 0.438) for TLOVR population that censored patients who withdrew for reasons other than virologic failure. Currier J et al. Ann Intern Med 2010;153:349-357 2010 by American College of Physicians
It is preferable to conduct adequately powered studies that will yield knowledge that will help both men and women than to do more studies that systematically provide poorer information for more than half the population. -Nancy E Adler, JAMA, December 22/29, 2010
Generalizability of Results: A Modeling Approach
Summary Major advances in HIV therapeutics have made it more difficult to evaluate the role of newer agents Non-inferiority designs are important in many areas of therapeutics- beware of pitfalls Composite endpoint trials are attractive but need to be reported properly Sex differences in outcome are important across all therapeutic areas